Dynamic Reconstruct for Network Photograph Exploration

Photograph sear reconstructmethods usually fail to capture the user’s intention when the query termism ambiguous. Therefore, reconstruct with user interactions, or active reconstruct, is highly demanded to effect veryimprove the search performance. The essential problem in active reconstruct is how to target the user’s intention. To complete this goal, this paper presents a structural information based sample selection strategy to reduce the user’s labeling efforts. Furthermore, to localize the user’s intention in the visual feature e space, a novel local-global discriminative dimension reduction algorithmic proposed. In this algorithm, a sub manifold is leer need by transferring the local geometry and the discriminate vet information from the labeled photographs to the whole (global) photographdatabase. Experiments on both synthetic datasets and a real Network photograph sear chdatasetdemenstruate he effectiveness of the proposed active reconstruct scheme, including both the structural information based active sample selection strategy and the local-global discriminative dimension reduction algorithm.